Child Doc 1: OGTT & ITT Data

2.a. Overview

• Each mouse received an OGTT and an ITT at baseline and endpoint. Therefore, we can assess the effects of hormone treatment via:
  • Intragroup / intramouse longitudinal changes from baseline to endpoint
  • Cross-sectional differences between groups at endpoint
• Useful group comparisons:
  • We can compare metabolic test results in hormone-exposed mice with results in saline-exposed mice within the same microbiome condition (Non-ABX or ABX) to quantify the metabolic effects of each hormone under a given gut microbiome environment.
  • We can compare metabolic test results in mice exposed to the same hormone but under different microbiome conditions to isolate any additive or interaction effects of hormones and gut microbiome presence on glucose homeostasis.
• For each test, we can look at:
  • The response over time (typical curve) in absolute numbers
  • Area under the curve (AUC)
  • The response over time as a % change from the fasting BG
    • In the literature, this is more commonly shown for ITTs than OGTTs
  • Fasting blood glucose from both tests (2 baseline and 2 endpoint measurements per mouse)

Here’s a simple review of the difference between what OGTTs and ITTs tell us:

• OGTT: Measures how efficiently an organism processes glucose, assessing glucose tolerance and the overall function of glucose regulatory mechanisms. A mouse’s response to an OGTT depends on both the ability of the pancreas to secrete insulin in response to glucose and the effectiveness of tissues (e.g., muscle, liver, adipose tissue) in taking up glucose out of circulation. Poor glucose processing during an OGTT suggests glucose intolerance.

• ITT: Measures how effectively insulin lowers blood glucose levels (i.e. the sensitivity / responsiveness of tissues to insulin). Reduced glucose clearance during an ITT suggests insulin resistance. Together, OGTTs and ITTs provide us with a complementary look at glucose homeostasis.

• AUC is a quantitative measure of the total blood glucose exposure over time during an OGTT or ITT. It is calculated by summing the area under the glucose curve generated by plotting glucose levels against time (i.e. integration). AUC allows for easy comparison between groups or treatments by summarizing complex time-series data into a single, interpretable metric.

NOTES:

• Two glucose measurements were taken for each mouse at each time point. If the 2 measurements were >25 apart, a 3rd measurement was taken. Mean blood glucose (of the 2-3 measurements) was calculated. Measurements >1 SD away from the mean were removed, and the mean was re-calculated.

• While OGTTs have a nearly 100% successful execution rate, ITTs occasionally are unsuccessful (likely due to human error in injection). For this experiment, I set a cutoff for a successful ITT as a >15% drop in blood glucose levels from fasting (time 0) to time 15. Data from tests that failed to meet this threshold were not included in the final data files. Unfortunately, this means that some mice do not have both baseline and endpoint ITT measurements.

2.a.i. OGTT and ITTs at baseline

While less relevant to the hypotheses of this project, it’s notable that there are differences between responses to the tolerance tests in abx and non-abx mice prior to any hormone injections (at baseline).

Absolute Responses

Here are two visualizations of those baseline differences. The first shows the mean response with standard error bars, and the second shows each individual mouse’s response with the group means overlaid:

Below is the same two visualization styles, but for ITT responses:

AUC

Now let’s look at the AUCs for the baseline OGTT and ITTs in ABX and Non-ABX mice:

Is there a significant difference at baseline between ABX and non-ABX OGTT AUCs?

## [1] "The t-test result is significant. p = 3.17e-06 ."

Is there a significant difference at baseline between ABX and non-ABX ITT AUCs?

## [1] "The t-test result is not significant. p = 0.962 ."

Key takeaway: We are observing discordance between the OGTT and ITT data. The OGTT data shows greater glucose tolerance (indicating lower insulin resistance) in the ABX mice (as expected), while in the ITT data, the Non-ABX mice exhibit greater glucose drops (potentially indicating lower insulin resistance). This apparent contradiction may be resolved by looking at both in terms of % change in addition to absolute change:

Percent Change Responses

NOTE: I’m unsure if it would make sense to calculate AUCs for the %-based graphs. There’s certainly an argument for it, but I don’t think I’ve ever seen it.

Fasting BG

In both tests, there is a clear pattern of lower fasting BG in ABX mice. This aligns with our previous findings. We can combine the fasting BG levels at both tests (since both are baselines, just a day apart), and run an LME to test the significance of these fasting differences.

## [1] "The linear mixed-effects model shows that the effect is significant. p = 1.98e-05 ."

As expected, there is a significant effect of ABX-status on fasting blood glucose levels (p = 0.00002).

2.a.ii. Longitudinal changes in OGTTs and ITTs in each group

Absolute Response Over Time

OGTTs Each hormone longitudinally in mice with a conventional gut microbiome:

AUC

Are any of these differences statistically significant? Since this is longitudinal changes, I can run paired t-tests.

1. PGH Non ABX Baseline to Endpoint OGTT:

## [1] "The t-test result is not significant. p = 0.0646 ."

2. PL Non ABX Baseline to Endpoint OGTT:

## [1] "The t-test result is significant. p = 0.013 ."

3. Saline Non ABX Baseline to Endpoint OGTT:

## [1] "The t-test result is not significant. p = 0.951 ."

---

Are any of these differences statistically significant? Not that while for the OGTTs, I could run paired t-tests, since some ITTs failed, I can only run unpaired t-tests.

1. PGH Non ABX Baseline to Endpoint ITT:

## [1] "The t-test result is significant. p = 0.0437 ."

2. PL Non ABX Baseline to Endpoint ITT:

## [1] "The t-test result is not significant. p = 0.995 ."

3. Saline Non ABX Baseline to Endpoint ITT:

## [1] "The t-test result is not significant. p = 0.111 ."

---

4. PGH ABX Baseline to Endpoint OGTT:

## [1] "The t-test result is not significant. p = 0.133 ."

5. PL ABX Baseline to Endpoint OGTT:

## [1] "The t-test result is significant. p = 0.00985 ."

6. Saline ABX Baseline to Endpoint OGTT:

## [1] "The t-test result is not significant. p = 0.297 ."

---

4. PGH ABX Baseline to Endpoint ITT:

## [1] "The t-test result is not significant. p = 0.752 ."

5. PL ABX Baseline to Endpoint ITT:

## [1] "The t-test result is significant. p = 0.0133 ."

6. Saline ABX Baseline to Endpoint ITT:

## [1] "The t-test result is not significant. p = 0.366 ."

Percent Change Response Over Time

Fasting BG

In each of the six groups, I want to see if there is a longitudinal change in fasting BG.

• Similar to what I did for baseline, I can combine the OGTT and ITT fasting BG values and run LMEs with mouse ID as a random effect (i.e., 4 values / mouse, 2 at each time point) to increase my power to detect longitudinal changes.

• Note that particularly in the ABX mice, I predict that fasting BG may decrease in all groups, including saline, because even though they are on ABX at both timepoints, there may be an additive effect of longer-term ABX exposure on BG levels.

• I’m actually not sure by looking at the figures that any will be significant, but we’ll see!

  1. PGH Non-ABX Baseline to Endpoint:

## [1] "The linear mixed-effects model shows that the effect is not significant. p = 0.67 ."

2. PL Non-ABX Baseline to Endpoint:

## [1] "The linear mixed-effects model shows that the effect is not significant. p = 0.304 ."

3. Saline Non-ABX Baseline to Endpoint:

## [1] "The linear mixed-effects model shows that the effect is not significant. p = 0.254 ."

4. PGH ABX Baseline to Endpoint:

## [1] "The linear mixed-effects model shows that the effect is not significant. p = 0.623 ."

5. PL ABX Baseline to Endpoint:

## [1] "The linear mixed-effects model shows that the effect is not significant. p = 0.282 ."

6. Saline ABX Baseline to Endpoint:

## [1] "The linear mixed-effects model shows that the effect is not significant. p = 0.649 ."

So NO groups saw significant changes in fasting BG from baseline to endpoint. Despite some clear changes in tolerance test responses. That’s interesting!

---

2.a.iii. Endpoint OGTTs and ITTs in each hormone group vs saline

Absolute Responses

Endpoint OGTTs in each hormone group:

We can include all 3 hormone groups together, or to make seeing patterns visually easier, separate each hormone to just be vs saline

Endpoint ITTs in each hormone group:

AUC

Now we can test the stats for each of these AUC pairings.

Non-ABX

PGH vs Saline

OGTT

## [1] "The t-test result is not significant. p = 0.692 ."

ITT

## [1] "The t-test result is not significant. p = 0.627 ."

PL vs Saline

OGTT

## [1] "The t-test result is not significant. p = 0.465 ."

ITT

## [1] "The t-test result is not significant. p = 0.472 ."

---

ABX

PGH vs Saline

OGTT

## [1] "The t-test result is not significant. p = 0.803 ."

ITT

## [1] "The t-test result is not significant. p = 0.869 ."

PL vs Saline

OGTT

## [1] "The t-test result is not significant. p = 0.135 ."

ITT

## [1] "The t-test result is significant. p = 0.0475 ."

---

Percent Change Responses

Fasting BG

Now let’s see if there are statistically significant differences between the hormone groups in endpoint fasting BG values.

• Similar to what I did for the prior fasting BG analyses, I will combine the OGTT and ITT fasting BG values and run LMEs with mouse ID as a random effect (i.e., 2 values / mouse) to increase my power to detect group differences.

Non-ABX

PGH vs Saline

## [1] "The linear mixed-effects model shows that the effect is significant. p = 0.00257 ."

PL vs Saline

## [1] "The linear mixed-effects model shows that the effect is significant. p = 0.0497 ."

ABX

PGH vs Saline

## [1] "The linear mixed-effects model shows that the effect is significant. p = 0.0294 ."

PL vs Saline

## [1] "The linear mixed-effects model shows that the effect is not significant. p = 0.772 ."

2.a.iv. Endpoint OGTTs and ITTs in each hormone group abx vs non-abx

Absolute Responses

AUC

Now we will test the significance for each antibiotic group pairing. I’m guessing these may all be significant.

PGH ABX vs Non ABX

OGTT

## [1] "The t-test result is significant. p = 0.0246 ."

ITT

## [1] "The t-test result is significant. p = 0.00461 ."

PL ABX vs Non ABX

OGTT

## [1] "The t-test result is significant. p = 0.00881 ."

ITT

## [1] "The t-test result is significant. p = 0.00882 ."

Saline ABX vs Non ABX

OGTT

## [1] "The t-test result is significant. p = 0.000758 ."

ITT

## [1] "The t-test result is significant. p = 0.0351 ."

As expected, for all 3 hormone groups, there is a signif dif in AUC for both OGTT and ITT between mice with and without a normal gut microbiome.

Percent Change Response

Basically, it looks like the metabolic differences between ABX and Non-ABX groups are largely driven by changes in fasting BG, so converting to % change makes them go away. But the metabolic differences between hormone groups are NOT driven by fasting BG, so they largely remain or sometimes magnify upon converting to % change.

Fasting BG

Now let’s see if there are statistically significant differences between the abx groups for each hormone in endpoint fasting BG values. Based on what we’ve seen so far, we can expect them all to be significant.

• Similar to what I did for the prior fasting BG analyses, I will combine the OGTT and ITT fasting BG values and run LMEs with mouse ID as a random effect (i.e., 2 values / mouse) to increase my power to detect group differences.

PGH ABX vs Non ABX

## [1] "The linear mixed-effects model shows that the effect is significant. p = 0.000278 ."

PL ABX vs Non ABX

## [1] "The linear mixed-effects model shows that the effect is significant. p = 0.00312 ."

Saline ABX vs Non ABX

## [1] "The linear mixed-effects model shows that the effect is significant. p = 0.00287 ."

2.a.v. Summary Table of Significant Results